Rule-based Knowledge Representation using a Database

Selecting a suitable method and knowledge engineering tool for a rule-based problem may be influenced by the fact that an expert system shell can be expensive and, contrary to database systems, may not be part of the installed computing environment. We show how to use a database (ORACLE) to implement the functionality of an expert system shell by a simple rule language which is automatically translated into SQL. The SQL interpreter is then used to "execute" these rules. We have applied this method to process control, in particular to checking timing sequences in the CERN particle accelerator complex. Our experience gained during more than one year of use, showed that the initial limitations concerning the execution speed could be overcome by software optimisation and hardware upgrades. Finally we compare the inference engine of this rule system implemented in a database with that of standard production rule expert system shells.